Training supervised image synthesis models requires a critic to compare two images: the ground truth to the result. Yet, this basic functionality remains an open problem. A popular line of approaches uses the L1 (mean absolute error) loss, either in the pixel or the feature space of pretrained deep networks. However, we observe that these losses tend to produce overly blurry and grey images, and other techniques such as GANs need to be employed to fight these artifacts. In this work, we introduce an information theory based approach to measuring similarity between two images. We argue that a good reconstruction should have high mutual information with the ground truth. This view enables learning a lightweight critic to "calibrate" a feature space in a contrastive manner, such that reconstructions of corresponding spatial patches are brought together, while other patches are repulsed. We show that our formulation immediately boosts the perceptual realism of output images when used as a drop-in replacement for the L1 loss, with or without an additional GAN loss.
Image registration aims to establish spatial correspondence across pairs, or groups of images, and is a cornerstone of medical image computing and computer-assisted-interventions. Currently, most deep learning-based registration methods assume that the desired deformation fields are globally smooth and continuous, which is not always valid for real-world scenarios, especially in medical image registration (e.g. cardiac imaging and abdominal imaging). Such a global constraint can lead to artefacts and increased errors at discontinuous tissue interfaces. To tackle this issue, we propose a weakly-supervised Deep Discontinuity-preserving Image Registration network (DDIR), to obtain better registration performance and realistic deformation fields. We demonstrate that our method achieves significant improvements in registration accuracy and predicts more realistic deformations, in registration experiments on cardiac magnetic resonance (MR) images from UK Biobank Imaging Study (UKBB), than state-of-the-art approaches.
Understanding the internal physiological changes accompanying the aging process is an important aspect of medical image interpretation, with the expected changes acting as a baseline when reporting abnormal findings. Deep learning has recently been demonstrated to allow the accurate estimation of patient age from chest X-rays, and shows potential as a health indicator and mortality predictor. In this paper we present a novel comparative study of the relative performance of radiologists versus state-of-the-art deep learning models on two tasks: (a) patient age estimation from a single chest X-ray, and (b) ranking of two time-separated images of the same patient by age. We train our models with a heterogeneous database of 1.8M chest X-rays with ground truth patient ages and investigate the limitations on model accuracy imposed by limited training data and image resolution, and demonstrate generalisation performance on public data. To explore the large performance gap between the models and humans on these age-prediction tasks compared with other radiological reporting tasks seen in the literature, we incorporate our age prediction model into a conditional Generative Adversarial Network (cGAN) allowing visualisation of the semantic features identified by the prediction model as significant to age prediction, comparing the identified features with those relied on by clinicians.
Collaborative representation-based classification (CRC) has demonstrated remarkable progress in the past few years because of its closed-form analytical solutions. However, the existing CRC methods are incapable of processing the nonlinear variational information directly. Recent advances illustrate that how to effectively model these nonlinear variational information and learn invariant representations is an open challenge in the community of computer vision and pattern recognition To this end, we try to design a new algorithm to handle this problem. Firstly, the second-order statistic, i.e., covariance matrix is applied to model the original image sets. Due to the space formed by a set of nonsingular covariance matrices is a well-known Symmetric Positive Definite (SPD) manifold, generalising the Euclidean collaborative representation to the SPD manifold is not an easy task. Then, we devise two strategies to cope with this issue. One attempts to embed the SPD manifold-valued data representations into an associated tangent space via the matrix logarithm map. Another is to embed them into a Reproducing Kernel Hilbert Space (RKHS) by utilizing the Riemannian kernel function. After these two treatments, CRC is applicable to the SPD manifold-valued features. The evaluations on four banchmarking datasets justify its effectiveness.
The Grad-CAM algorithm provides a way to identify what parts of an image contribute most to the output of a classifier deep network. The algorithm is simple and widely used for localization of objects in an image, although some researchers have point out its limitations, and proposed various alternatives. One of them is Grad-CAM++, that according to its authors can provide better visual explanations for network predictions, and does a better job at locating objects even for occurrences of multiple object instances in a single image. Here we show that Grad-CAM++ is practically equivalent to a very simple variation of Grad-CAM in which gradients are replaced with positive gradients.
We introduce ArtBench-10, the first class-balanced, high-quality, cleanly annotated, and standardized dataset for benchmarking artwork generation. It comprises 60,000 images of artwork from 10 distinctive artistic styles, with 5,000 training images and 1,000 testing images per style. ArtBench-10 has several advantages over previous artwork datasets. Firstly, it is class-balanced while most previous artwork datasets suffer from the long tail class distributions. Secondly, the images are of high quality with clean annotations. Thirdly, ArtBench-10 is created with standardized data collection, annotation, filtering, and preprocessing procedures. We provide three versions of the dataset with different resolutions ($32\times32$, $256\times256$, and original image size), formatted in a way that is easy to be incorporated by popular machine learning frameworks. We also conduct extensive benchmarking experiments using representative image synthesis models with ArtBench-10 and present in-depth analysis. The dataset is available at https://github.com/liaopeiyuan/artbench under a Fair Use license.
Pathological diagnosis relies on the visual inspection of histologically stained thin tissue specimens, where different types of stains are applied to bring contrast to and highlight various desired histological features. However, the destructive histochemical staining procedures are usually irreversible, making it very difficult to obtain multiple stains on the same tissue section. Here, we demonstrate a virtual stain transfer framework via a cascaded deep neural network (C-DNN) to digitally transform hematoxylin and eosin (H&E) stained tissue images into other types of histological stains. Unlike a single neural network structure which only takes one stain type as input to digitally output images of another stain type, C-DNN first uses virtual staining to transform autofluorescence microscopy images into H&E and then performs stain transfer from H&E to the domain of the other stain in a cascaded manner. This cascaded structure in the training phase allows the model to directly exploit histochemically stained image data on both H&E and the target special stain of interest. This advantage alleviates the challenge of paired data acquisition and improves the image quality and color accuracy of the virtual stain transfer from H&E to another stain. We validated the superior performance of this C-DNN approach using kidney needle core biopsy tissue sections and successfully transferred the H&E-stained tissue images into virtual PAS (periodic acid-Schiff) stain. This method provides high-quality virtual images of special stains using existing, histochemically stained slides and creates new opportunities in digital pathology by performing highly accurate stain-to-stain transformations.
In this work, we study the problem of non-blind image deconvolution and propose a novel recurrent network architecture that leads to very competitive restoration results of high image quality. Motivated by the computational efficiency and robustness of existing large scale linear solvers, we manage to express the solution to this problem as the solution of a series of adaptive non-negative least-squares problems. This gives rise to our proposed Recurrent Least Squares Deconvolution Network (RLSDN) architecture, which consists of an implicit layer that imposes a linear constraint between its input and output. By design, our network manages to serve two important purposes simultaneously. The first is that it implicitly models an effective image prior that can adequately characterize the set of natural images, while the second is that it recovers the corresponding maximum a posteriori (MAP) estimate. Experiments on publicly available datasets, comparing recent state-of-the-art methods, show that our proposed RLSDN approach achieves the best reported performance both for grayscale and color images for all tested scenarios. Furthermore, we introduce a novel training strategy that can be adopted by any network architecture that involves the solution of linear systems as part of its pipeline. Our strategy eliminates completely the need to unroll the iterations required by the linear solver and, thus, it reduces significantly the memory footprint during training. Consequently, this enables the training of deeper network architectures which can further improve the reconstruction results.
Convolutional neural networks (CNNs) have shown exceptional performance for a range of medical imaging tasks. However, conventional CNNs are not able to explain their reasoning process, therefore limiting their adoption in clinical practice. In this work, we propose an inherently interpretable CNN for regression using similarity-based comparisons (INSightR-Net) and demonstrate our methods on the task of diabetic retinopathy grading. A prototype layer incorporated into the architecture enables visualization of the areas in the image that are most similar to learned prototypes. The final prediction is then intuitively modeled as a mean of prototype labels, weighted by the similarities. We achieved competitive prediction performance with our INSightR-Net compared to a ResNet baseline, showing that it is not necessary to compromise performance for interpretability. Furthermore, we quantified the quality of our explanations using sparsity and diversity, two concepts considered important for a good explanation, and demonstrated the effect of several parameters on the latent space embeddings.
What can we learn about a scene by watching it for months or years? A video recorded over a long timespan will depict interesting phenomena at multiple timescales, but identifying and viewing them presents a challenge. The video is too long to watch in full, and some occurrences are too slow to experience in real-time, such as glacial retreat. Timelapse videography is a common approach to summarizing long videos and visualizing slow timescales. However, a timelapse is limited to a single chosen temporal frequency, and often appears flickery due to aliasing and temporal discontinuities between frames. In this paper, we propose Video Temporal Pyramids, a technique that addresses these limitations and expands the possibilities for visualizing the passage of time. Inspired by spatial image pyramids from computer vision, we developed an algorithm that builds video pyramids in the temporal domain. Each level of a Video Temporal Pyramid visualizes a different timescale; for instance, videos from the monthly timescale are usually good for visualizing seasonal changes, while videos from the one-minute timescale are best for visualizing sunrise or the movement of clouds across the sky. To help explore the different pyramid levels, we also propose a Video Spectrogram to visualize the amount of activity across the entire pyramid, providing a holistic overview of the scene dynamics and the ability to explore and discover phenomena across time and timescales. To demonstrate our approach, we have built Video Temporal Pyramids from ten outdoor scenes, each containing months or years of data. We compare Video Temporal Pyramid layers to naive timelapse and find that our pyramids enable alias-free viewing of longer-term changes. We also demonstrate that the Video Spectrogram facilitates exploration and discovery of phenomena across pyramid levels, by enabling both overview and detail-focused perspectives.